JP2010129025A - Word relation determining device and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To determine the relation regarding higher and lower order relations of verbs, with high accuracy. <P>SOLUTION: The word relation determining device 10 sets a first verb and a second verb as an object to be determined; obtains a first name group including one or more nouns, which coincides/coincide with a given condition about the first verb from a plurality of sentences; obtains a second noun group including one or more nouns, which coincides/coincide with a given condition about the second verb from the plurality of sentences; and determines whether having superior and inferior relations between the first verb and the second verb, on the basis of comparison of a variation in the first noun group with a variation in the second noun group in a noun thesaurus in which a plurality of names are represented in a tree structure, on the basis of superior and inferior relations of the concept belonging to each noun. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

  The present invention relates to a word relationship determination device and a program.

  A thesaurus that describes hierarchical relationships based on the concept (meaning) between words plays an important role in supplementing and expanding input keywords when searching for text, and normalizing keywords when sorting documents. Language information.

  The thesaurus is generally created manually, but the construction of the thesaurus requires high consistency and consistency, and the human work requires a large processing load, and there is a technology for mechanically constructing the thesaurus. It has been needed.

Therefore, as described in Patent Document 1 below, the distance between the two words is calculated based on the frequency of words appearing in the vicinity of each of the two words, and clustering based on the calculated distance is performed to obtain a thesaurus. Some have proposed a technology for mechanically constructing the system.
JP 2000-231572 A

  When constructing a thesaurus for verbs, words that appear in the vicinity of verbs that are in a higher-order and lower-order relationship often have a very close frequency distribution, and words that appear in the vicinity of the verb If it is performed based on the frequency distribution, the determination may not be made with high accuracy.

  One of the objects of the present invention is to provide a word relationship determination device and a program capable of accurately determining upper and lower relationships for a verb.

  In order to achieve the above object, the invention of a word relationship determination device according to claim 1 is characterized in that setting means for setting a first verb and a second verb as determination targets, and the first verb from a plurality of sentences. A first noun group obtaining means for obtaining a first noun group including one or more nouns that match a given condition with respect to the second verb from the plurality of sentences, the first noun group obtaining means being matched with the given condition A second noun group acquiring means for acquiring a second noun group including one or a plurality of nouns; and the noun thesaurus in which the plurality of nouns are represented in a tree structure based on the upper and lower relations of the concept of each noun. Determination means for determining whether or not the first verb and the second verb are in an upper-lower relationship based on a comparison between the variation of the first noun group and the variation of the second noun group. It is characterized by including these.

  According to a second aspect of the present invention, in the word relationship determination device according to the first aspect, the determination means is predetermined for the variation of the first noun group and the variation of the second noun group. In the case where there is more than the difference, it is determined that the verb related to the one with the larger variation is in the upper concept and the verb related to the other is in the lower concept.

  The invention according to claim 3 is the word relationship determination device according to claim 2, wherein the given condition includes a plurality of conditions, and the determination means is based on any of the plurality of conditions. If there is more than a predetermined difference between the variation of the first noun group and the variation of the second noun group obtained in this way, the verb related to the larger variation is in the higher order and the verb related to the other is in the lower relationship. It is determined that it exists.

  According to a fourth aspect of the present invention, in the word relationship determination device according to any one of the first to third aspects, the given condition is a relationship between a verb set as a target by the setting means and a dependency. At least one of the following conditions: a noun condition in which the noun appears in the vicinity of the verb set for the object, and a noun condition included in the phrase through a representation of a causal relationship to the verb set for the object It is characterized by including one.

  Further, the invention according to claim 5 is the word relationship determination device according to any one of claims 1 to 4, wherein the variation of the noun group in the noun thesaurus is the number of nouns included in the noun group. It is calculated based on the number of hops in the noun thesaurus.

  The invention according to claim 6 is the word relationship determination device according to any one of claims 1 to 5, wherein the first verb and the second verb are based on the first and second noun groups. And a similarity determination unit that determines similarity, and when the similarity determination unit determines that the similarity is similar, the determination unit has a high-level and low-level relationship between the first verb and the second verb It is characterized by determining whether or not.

  The invention described in claim 7 is the word relationship determination device according to claim 6, wherein the similarity determination unit determines each noun included in the first and second noun groups in advance in the noun thesaurus. The concept is converted into a predetermined reference concept, and the similarity between the first verb and the second verb is determined based on the frequency of the reference concept included in the converted first and second noun groups. It is characterized by that.

  Further, the invention according to claim 8 is the word relationship determination device according to claim 6 or 7, wherein the similarity determination unit determines that the similarity is similar, and the determination unit determines that there is no upper or lower relationship. The determined first verb and the second verb are determined to have a synonym relationship.

  According to another aspect of the present invention, there is provided a word relation determination device according to the present invention, wherein the first verb and the second verb are set as determination targets, and the expression representing the causal relationship from a plurality of sentences to the first verb. Via a first noun group acquisition means for acquiring a first noun group including one or more nouns included in the clause, and an expression representing a causal relationship from the plurality of sentences to the second verb Based on the comparison between the second noun group acquisition means for acquiring the second noun group including one or a plurality of nouns included in the phrase and the variation of the first noun group and the variation of the second noun group. And determining means for determining whether or not the first verb and the second verb are in a high-order and low-order relationship.

  According to a tenth aspect of the present invention, there is provided a program that sets a first verb and a second verb as determination targets, and that satisfies a given condition with respect to the first verb from a plurality of sentences. Or a first noun group acquisition step of acquiring a first noun group including a plurality of nouns, and a second noun that includes one or more nouns that match the given condition with respect to the second verb from the plurality of sentences. A second noun group acquisition step of acquiring a noun group of the first noun group in a noun thesaurus that represents a plurality of nouns in a tree structure based on the upper and lower relations of the concept of each noun and the noun group Determining whether or not the first verb and the second verb are in an upper-lower relationship based on the comparison with the variation of the second noun group. Features.

  The invention of the program according to claim 11 includes a setting step of setting the first verb and the second verb as determination targets, and an expression expressing a causal relationship from a plurality of sentences to the first verb. A first noun group acquisition step of acquiring a first noun group including one or a plurality of nouns included in the clause, and a phrase representing the causal relationship from the plurality of sentences to the second verb. Based on a second noun group acquisition step of acquiring a second noun group including one or a plurality of nouns included, and a comparison between the variation of the first noun group and the variation of the second noun group, A determination step of determining whether or not the first verb and the second verb are in an upper-lower relationship is executed by a computer.

  According to the inventions described in claims 1 and 10, the upper and lower determinations of each verb are accurately performed using the conceptual distribution of the noun group acquired for each of the first verb and the second verb. Yes.

  According to the second aspect of the present invention, it is possible to determine the broader conceptual distribution as higher and the other as lower.

  According to the third aspect of the present invention, even when the upper / lower determination cannot be made by one condition, the upper / lower relation between the verbs can be determined by another condition.

  According to invention of Claim 4, a noun group with a high relation with a verb is acquirable.

  According to the invention described in claim 5, it is possible to calculate by converting the variation between nouns into an objective distance in the noun thesaurus.

  According to the sixth aspect of the present invention, it is possible to determine that verbs having different conceptual distributions of similar and extracted noun groups are in a higher and lower order relationship.

  According to the seventh aspect of the present invention, the frequency analysis of the noun group obtained for the verb can be performed for each noun concept.

  According to invention of Claim 8, a synonym can be determined.

  According to the ninth and eleventh aspects of the present invention, it is possible to make upper / lower determinations between verbs based on a noun group having a causal relationship with a verb.

  DESCRIPTION OF EXEMPLARY EMBODIMENTS Hereinafter, preferred embodiments (hereinafter referred to as embodiments) for carrying out the invention will be described with reference to the drawings.

  In FIG. 1, the functional block diagram of the word relationship determination apparatus 10 which concerns on this embodiment was shown. As shown in FIG. 1, the word relationship determination device 10 includes a text data storage unit 12, a noun thesaurus storage unit 14, a determination target verb setting unit 16, a dependency noun extraction unit 18, a similarity determination unit 20, and a first determination. A unit 22, a causal noun extraction unit 24, and a second determination unit 26. The functions of the above-described units are such that a computer having a control unit such as a CPU, a storage unit such as a memory, and an input / output unit that transmits and receives data to and from an external device reads a program stored in a computer-readable information storage medium. It may be realized by executing. The program may be supplied to the word relationship determination apparatus 10 as a computer by an information storage medium, or may be supplied via a data communication network such as the Internet.

  The text data storage unit 12 includes a storage device such as a magnetic disk, and stores one or more document information including one or more sentences. The document information may be configured as a data file including character string data.

  The noun thesaurus storage unit 14 stores a noun thesaurus constructed for nouns, and the noun thesaurus is information represented by a tree structure in which nouns are hierarchized by upper and lower relationships. FIG. 2 shows an example of a noun thesaurus.

  As shown in FIG. 2, the noun thesaurus is a tree structure connected by a relationship of a superordinate concept and a subordinate concept relative to the superordinate concept sequentially from the root. In this embodiment, a part of the noun thesaurus is Selected as the standard concept. This reference concept may be selected individually for each concept, or may be selected according to the depth of the hierarchy from the root. Details of processing using the reference concept will be described later.

  The determination target verb setting unit 16 sets a set of verbs that are targets for determining the relationship between each other. Hereinafter, a pair set by the determination target verb setting unit 16 is referred to as a verb A and a verb B. The determination target verb setting unit 16 may accept an input from the user and set a determination target verb, or select a set of verbs from a list of verbs prepared in advance, and select the selected verb. It may be set as a determination target.

  The dependency noun extraction unit 18 extracts nouns having a dependency relationship with respect to each of the set of verbs set by the determination target verb setting unit 16. The dependency noun extraction unit 18 searches the text data stored in the text data storage unit 12 for a sentence including the verbs (verb A and verb B) set by the determination target verb setting unit 16, and the searched sentence After performing morphological analysis and syntactic analysis, the phrase related to the set verb is specified. Then, the dependency noun extraction unit 18 selects the case particle when the end of the phrase is “ga”, “wo”, “de”, or “ni” among the specified clauses. Extract the noun before. The dependency noun extraction unit 18 records the extracted noun along with the appearance frequency of the noun for each word of the case particle. The dependency noun extraction unit 18 performs the above processing for each of the verb A and the verb B.

  The similarity determination unit 20 determines whether the verb A and the verb B are similar. Here, similarity means that verb A and verb B have a synonym relationship. In the present embodiment, the similarity determination unit 20 uses all the nouns included in the noun group extracted by the dependency noun extraction unit 18 for the verb to be determined (verb A or verb B) as the noun thesaurus storage unit 14. Based on the noun thesaurus stored in, it is converted into a higher-order reference concept word. For example, according to the noun thesaurus shown in FIG. 2, the word “Bulldog” is converted to “animal” because there is a standard concept “animal” at the top. The similarity determination unit 20 converts the noun group obtained for each of the verb A and the verb B into the reference concept.

  The similarity determination unit 20 generates a feature vector for each of the verb A and the verb B based on the noun group (referred to as the reference concept noun group) converted into the reference concept word by the above processing. The feature vector generation processing in the present embodiment will be described below.

  First, reference concept noun groups extracted for the case particles “ga”, “wo”, “de”, and “ni” for the verb A are defined as GA1, GA2, GA3, and GA4, respectively. Here, GA1 = {wi | i = 1 to n}, where n is the number of different words included in GA1, and if the appearance frequency of wi is NA1wi, the appearance ratio RA1wi is the number of elements of GA1 is NA1. In this case, RA1wi = NA1wi / NA1 is calculated. For GA1, feature vector GA1 = (RA1w1, RA1w2,..., RA1wn). For GA2, GA3, and GA4, feature vectors are generated based on the same processing. Then, the feature vectors GA1, GA2, GA3, GA4 are concatenated to convert the feature vector RA of the verb A into RA = (RA1w1, RA1w2,..., RA1wn, RA2w1, RA2w2,..., RA2wn, RA3w1, RA3w2,. .., RA3wn, RA4w1, RA4w2,..., RA4wn). A feature vector RB is similarly generated for the verb B. RA and RB are each normalized to a vector length of 1, and feature vectors RA and RB are assumed to be normalized vectors.

  The similarity determination unit 20 calculates the inner product of the feature vector RA and the feature vector RB of each of the generated verb A and verb B, and classifies the verb A and the verb B based on the size of the calculated inner product. Determine no. Specifically, the similarity determination unit 20 determines that the verb A and the verb B are similar when the inner product RA · RB of the feature vector RA and the feature vector RB is larger than the threshold T1 (0 <T1 <1). It is determined that there is. The threshold value T1 may be determined in advance.

  The first determination unit 22 determines whether or not the verb A and the verb B set as the determination target are in a relationship of higher and lower concepts. When the similarity determination unit 20 determines that the verb A and the verb B to be determined are similar, the first determination unit 22 determines the upper and lower concepts. The first determination unit 22 first calculates the distribution distance on the noun thesaurus stored in the noun thesaurus storage unit 14 of each noun group extracted by the dependency noun extraction unit 18 for the verb A and the verb B. Based on the calculated distribution distance, the upper and lower relations of the verb A and the verb B are determined. Hereinafter, specific processing of the first determination unit 22 in the present embodiment will be described.

  First, the first determination unit 22 extracts a set of nouns from the set GA1 extracted for the case particle “ga”, and acquires a distance on the noun thesaurus of the extracted nouns. The distance on the noun thesaurus in the present embodiment is the number of hops from one noun to the other noun on the noun thesaurus. The number of hops is, for example, 1 if a noun is connected to another noun by a direct link, and 2 if it is connected via another noun. Specifically, in the noun thesaurus shown in FIG. 2, “Bulldog” and “Dog” have 1 hop, and “Bulldog” and “Cat” have 3 hops. The first determination unit 22 acquires the number of hops for all pairs of nouns included in GA1, and sets the average value of the number of hops as the distance MA1 on the noun thesaurus related to GA1. Similarly, the first determination unit 22 calculates distances MA2 to MA4 on the noun thesaurus for the other sets GA2 to GA4, and a vector MA = (MA1, MA2, MA3 indicating the distribution of the verb A on the noun thesaurus. , MA4). The first determination unit 22 performs the same process for the verb B, and obtains a vector MB indicating the distribution of the verb B on the noun thesaurus.

  Next, the first determination unit 22 calculates an inter-vector distance ‖MA-MB‖ between the vector MA and the vector MB, and when this value is larger than a predetermined threshold T2 (T2> 0), Verb A and verb B are determined to have a high-level and low-level relationship. If not, it is determined that the high-level and low-level relationship is unknown. When it is determined that there is an upper / lower relationship according to the above criteria, it is determined that the larger vector norm is the higher concept and the smaller vector is the lower concept. This is because the words of the higher concept are used more widely than the words of the lower concept, and the distribution is considered wide.

  The causal relation noun extraction unit 24, when the first determination unit 22 determines that the upper and lower relationship between the verb A and the verb B is unknown, the sentence including the verb A or the verb B is a text data storage unit. Sentences that are extracted from No. 12 and that have a causal relationship such as “because”, “because of”, “because of” or the like before the verb A or the verb B Search for. And the causal relation noun extraction part 24 extracts the noun contained in the phrase which appears before the expression showing the causal relation in the searched sentence, and obtains a noun group (causal relation noun group). The causal relation noun extraction unit 24 extracts a causal relation noun group for each of verb A and verb B, and a set of causal relation noun groups extracted for verb A is a causal relation noun extracted for verb B. Let HB be the set of groups.

  The second determination unit 26 determines the upper and lower relationship between the verb A and the verb B again based on the causal relationship noun group extracted for each verb by the causal relationship noun extraction unit 24. Hereinafter, specific processing of determination performed by the second determination unit 26 will be described.

  The second determination unit 26 calculates the distribution distance on the noun thesaurus of the causal noun group HA extracted for the verb A. The second determination unit 26 acquires the number of hops on the noun thesaurus for all pairs of nouns included in the HA, and sets the average value of the acquired hop counts as the distribution distance LA on the noun thesaurus of the HA. calculate. The second determination unit 26 calculates the distribution distance LB on the noun thesaurus of HB in the same manner for the set HB of causal noun groups extracted for the verb B. Then, the second determination unit 26 obtains the difference between LA and LB, and if the obtained difference is larger than a predetermined threshold T3 (T3> 0), the second and second determination units 26 are higher and lower in verb A and verb B. Otherwise, it is determined that verb A and verb B are synonyms. If it is determined by the above criteria that there is an upper / lower relationship, it is determined that the one with the larger distribution distance is the upper concept and the one with the smaller distribution distance is the lower concept.

  The word relationship determination apparatus 10 is based on the above-described three determination results of the similarity determination unit 20, the first determination unit 22, and the second determination unit 26. Determine. The word relationship determination device 10 sequentially executes the same determination process for other verb pairs, thereby obtaining a concept of a higher and lower order of a plurality of verbs and constructing a verb thesaurus.

  Next, referring to the flowchart shown in FIG. 3, the flow of the determination process of the upper and lower concepts of the verb by the word relationship determination device 10 will be described.

  As shown in FIG. 3, the word relationship determination device 10 receives an input of a set of verbs to be determined (S101), and determines the dependency relationship for each set of verbs received from stored text data. A noun group is extracted (S102). Next, the word relationship determination device 10 converts each noun of the extracted noun group into a noun thesaurus reference concept (S103), and generates a feature vector for each verb based on the converted reference concept word. The similarity of the target verb set is determined (S104).

  When determining that the verbs are similar (S104: Y), the word relationship determination device 10 further calculates the distribution distance in the noun thesaurus of the noun group extracted for each verb (S105). The word relationship determination device 10 determines whether or not the calculated distribution distance is greater than or equal to a threshold value (S106). If it is determined that the calculated distribution distance is equal to or greater than the threshold value (S106: Y), It is determined as lower (S107). In addition, when determining that the word relationship determination apparatus 10 is less than the threshold (S106: N), the word relationship determination device 10 extracts a causal relationship noun group including nouns related to each of the determination target verbs using an expression representing the causal relationship (S108). ), It is determined whether or not the distribution distance in the noun thesaurus of the extracted causal noun group is greater than or equal to a threshold (S109). Here, when it is determined that the threshold value is greater than or equal to the threshold (S109: Y), the one with the larger distribution distance is determined to be higher and the other is determined to be lower (S107). On the other hand, when it is determined that it is less than the threshold (S109: N), it is determined that the verb is a synonym (S110). If it is determined in S104 that they are not similar (S104: N), it is determined that the verb pair does not have a higher-lower relationship (S111), and the process ends.

  Next, as specific examples, (1) “permit” and “read”, (2) “permit” and “authorize”, and (3) “read” and “read carefully”, word relationships for three examples The example of determination by the determination apparatus 10 is shown.

  FIGS. 4A to 4D show examples of nouns having a dependency relationship with respect to each verb extracted from the text data and the appearance frequency of the nouns. 4A, FIG. 4B, FIG. 4C, and FIG. 4D show the nouns and appearance frequencies extracted for “Allow”, “Read”, “Authorize”, and “Read carefully”, respectively.

  5A to 5D show tables showing the frequency of words of each reference concept for each case particle when the noun extracted for each verb is converted to the reference concept. FIG. 5A, FIG. 5B, FIG. 5C, and FIG. 5D show tables for “permit”, “read”, “permit”, and “peruse”, respectively.

  First, (1) determination of the relationship between “permitted” and “read” will be described. As shown in FIG. 5A and FIG. 5B, in “permit” and “read”, the words of the reference concept obtained for the case particle “ga” for “permit” are “region”, “group”, “ “Role” and the like have a high appearance frequency, and the case particle “O” has a high appearance frequency such as “operation” and “production”. On the other hand, the words of the basic concept obtained for the case particle “ga” for “read” are concentrated in the “person”, and the frequency of occurrence in the “stationery” is concentrated for the case particle “ha”. In the case of “permit” and “read”, the frequency distribution of the words of the standard concept becomes large. Therefore, the inner product value of the feature vectors based on the frequency distribution of the words of the reference concept for “permit” and “read” becomes small, and it is determined that there is no similarity. Therefore, the upper and lower determinations are not performed for “permit” and “read”.

  Next, (2) determination of the relationship between “permitted” and “permitted” will be described. As shown in FIG. 5A and FIG. 5C, the reference words obtained for the case particle “ga” for both verbs have a high frequency of appearance of “region”, “group”, “role”, etc. ”Has a common tendency that the appearance frequency of“ operation ”,“ production ”, etc. is high. Therefore, the inner product value of the feature vectors based on the frequency distribution of the words of the reference concept for “permit” and “permit” is increased, and it is determined that there is similarity.

  Therefore, the first determination unit 22 calculates the average distance in the noun thesaurus for each noun shown in FIG. 4A and FIG. 4C for “permit” and “permit”, and determines the upper and lower relationship between the two. To do. Specifically, “permit” and “permit” tend to be used for administration, etc., whereas “permit” is more widely used. The average distance in the noun thesaurus is greater for “permit” than for “permit”. Therefore, it is determined that “permit” and “authorize” have a high-order relationship, “permit” has a high-order relationship, and “permit” has a low-order relationship.

  Next, (3) the determination of the relationship between “read” and “read carefully” will be described. As shown in FIGS. 5B and 5D, both have the same frequency distribution for case particles, and the similarity determination unit 20 determines that both are similar. And in the 1st determination part 22, since there is no big difference in distribution in both noun groups also about the average distance in the noun thesaurus of the noun group extracted about both, a difference does not appear also in the calculated average distance, both It is determined that the relationship between the upper and lower levels is unknown.

  Therefore, the causal relation noun extraction unit 24 extracts sentences including causal relation expressions such as “~ so”, “~ for”, and “for reason” among sentences including “read” and “read carefully”. In the extracted sentence, nouns that appear before the causal relation expression are further extracted. For example, with regard to “reading”, the sentences “reading a reference book for exams”, “reading a book to gain knowledge”, and “reading a book for leisure” were extracted. Then, a causal noun group of “examination”, “knowledge”, and “free time” is obtained from the sentence thus extracted. On the other hand, with regard to “read carefully”, for example, “reading a reference book for an examination”, “reading a textbook for a test”, and “reading a book carefully because it requires memorization” If extracted, a causal noun group of “test”, “test”, and “memorization” is obtained from the sentence thus extracted. The second determination unit 26 calculates the average distance in the noun thesaurus of the causal noun group obtained for each verb. In the above example, the cause of the act of “reading carefully” is “test”, “test”, etc. Concentrating on nouns in some semantic categories, and on the other hand, “reading” has various causes of action, so the average distance in the noun thesaurus of both causal noun groups is “reading” is better than “reading”. Larger than Accordingly, “read” and “read carefully” are determined by the second determination unit 26 that “read” is higher and “read” is lower.

  The present invention is not limited to the above-described embodiment. For example, for a set of verbs determined to be similar by the similarity determination unit 20, only the determination by the second determination unit 26 is performed, and upper and lower levels are determined. It is also possible to make a determination. In addition, the similarity determination unit 20 and the first determination unit 22 make the determination using a noun that has a dependency relationship with the determination target verb extracted by the dependency noun extraction unit 18. The noun may be a noun having a co-occurrence relationship in the sentence with the verb to be determined. Further, in the first determination unit 22 and the second determination unit 26, the upper and lower determinations are made based on the commonality of the elements for the noun groups extracted by the dependency noun extraction unit 18 and the causal noun extraction unit 24, respectively. It is good also as determining which is high-order based on the inclusion relationship of both noun groups. In addition, when searching for verbs that do not exist in the verb thesaurus from the text data, and determining each verb in the verb thesaurus for the searched verb, and when the upper and lower, synonyms are determined, The searched verb may be added to the verb thesaurus.

It is a functional block diagram of the word relationship determination apparatus which concerns on this embodiment. It is a figure which shows an example of a noun thesaurus. It is a flowchart of the determination process of the superordinate concept of a verb. It is the figure which showed an example of the noun which has a dependency relation about "permit", and the appearance frequency of the noun. It is the figure which showed an example of the noun which has a dependency relation about "reading", and the appearance frequency of the noun. It is the figure which showed an example of the noun which has a dependency relation about "authorize", and the appearance frequency of the noun. It is the figure which showed an example of the noun which has a dependency relation about "read carefully", and the appearance frequency of the noun. It is a figure which shows the table which represented the frequency of the word of each reference | standard concept about "permit" for every case particle. It is a figure which shows the table which represented the frequency of the word of each reference | standard concept about "reading" for every case particle. It is a figure which shows the table which represented the frequency of the word of each reference | standard concept about "permit" for every case particle. It is a figure which shows the table which represented the frequency of the word of each reference | standard concept about "reading carefully" for every case particle.

Explanation of symbols

  DESCRIPTION OF SYMBOLS 10 Word relationship determination apparatus, 12 Text data storage part, 14 Noun thesaurus storage part, 16 Determination object verb setting part, 18 Dependency noun extraction part, 20 Similarity determination part, 22 1st determination part, 24 Causal relation noun extraction part , 26 Second determination unit.

Claims (11)

Setting means for setting the first verb and the second verb as determination targets;
First noun group acquisition means for acquiring a first noun group including one or more nouns that match a given condition with respect to the first verb from a plurality of sentences;
Second noun group acquisition means for acquiring a second noun group including one or more nouns that match the given condition with respect to the second verb from the plurality of sentences;
Based on a comparison between the variation of the first noun group and the variation of the second noun group in a noun thesaurus that represents a plurality of nouns in a tree structure based on the upper and lower relations of the concept of each noun, Determining means for determining whether or not the first verb and the second verb are in a high-order and low-order relationship. When there is a difference greater than a predetermined difference between the variation of the first noun group and the variation of the second noun group, the determination unit is configured to use the verb related to the larger variation and the verb related to the other The word relationship determination device according to claim 1, wherein the word relationship determination device is determined to be in the concept. The given condition includes a plurality of conditions;
The determination means has a large variation when there is a difference greater than or equal to a predetermined value between the variation of the first noun group and the variation of the second noun group acquired based on any of the plurality of conditions. The word relationship determination device according to claim 2, wherein a verb related to the direction is determined to be in a higher order and a verb related to the other is determined to be in a lower relationship. The given condition is a condition of a noun that has a dependency relationship with a verb set as a target by the setting means, a condition of a noun that appears in the vicinity of a verb set as the target, and the target The word relationship determination device according to any one of claims 1 to 3, further comprising at least one condition of nouns included in the phrase through an expression representing a causal relationship with the verb. The word according to claim 1, wherein the variation of the noun group in the noun thesaurus is calculated based on the number of hops in the noun thesaurus of each set of nouns included in the noun group. Relationship determination device. A similarity determination means for determining similarity between the first verb and the second verb based on the first and second noun groups;
The determination unit determines whether the first verb and the second verb are in a high-order or low-order relationship when the similarity determination unit determines that the similarity is similar. The word relationship determination device according to any one of 1 to 5. The similarity determination unit converts each noun included in the first and second noun groups into a concept of a predetermined standard in the noun thesaurus, and converts the nouns into the converted first and second noun groups. The word relationship determination device according to claim 6, wherein the similarity between the first verb and the second verb is determined based on a frequency of a reference concept included. Determining that the first verb and the second verb determined to be similar by the similarity determination unit and not determined to be in a higher or lower relationship by the determination unit are in a synonym relationship. The word relationship determination device according to claim 6 or 7, wherein: Setting means for setting the first verb and the second verb as determination targets;
First noun group acquisition means for acquiring a first noun group including one or a plurality of nouns included in the phrase from a plurality of sentences via an expression representing a causal relationship to the first verb;
Second noun group acquisition means for acquiring a second noun group including one or a plurality of nouns included in the clause through an expression representing a causal relationship to the second verb from the plurality of sentences;
Judgment whether or not the first verb and the second verb are in an upper-lower relationship based on the comparison between the variation of the first noun group and the variation of the second noun group A word relationship determination apparatus characterized by comprising: means. A setting step for setting the first verb and the second verb as determination targets;
A first noun group obtaining step of obtaining a first noun group including one or more nouns that match a given condition with respect to the first verb from a plurality of sentences;
A second noun group obtaining step of obtaining a second noun group including one or more nouns that match the given condition with respect to the second verb from the plurality of sentences;
Based on a comparison between the variation of the first noun group and the variation of the second noun group in a noun thesaurus that represents a plurality of nouns in a tree structure based on the upper and lower relations of the concept of each noun, A program for causing a computer to execute a determination step of determining whether or not the first verb and the second verb are in an upper-lower relationship. A setting step for setting the first verb and the second verb as determination targets;
A first noun group acquisition step of acquiring a first noun group including one or a plurality of nouns included in the phrase through a representation representing a causal relationship to the first verb from a plurality of sentences;
A second noun group acquisition step of acquiring a second noun group including one or a plurality of nouns included in the phrase through an expression representing a causal relationship to the second verb from the plurality of sentences;
Judgment whether or not the first verb and the second verb are in an upper-lower relationship based on the comparison between the variation of the first noun group and the variation of the second noun group A program for causing a computer to execute the steps.

Images